Lecture: New Constructive Aspects of the Lovász Local Lemma, and their Applications June 15, 2011
|
|
- Bathsheba Spencer
- 6 years ago
- Views:
Transcription
1 Approximation Algorithms Workshop June 13-17, 2011, Princeton Lecture: New Constructive Aspects of the Lovász Local Lemma, and their Applications June 15, 2011 Aravind Srinivasan Scribe: Ioana Bercea 1 Introduction The Lovász Local Lemma (LLL) is a powerful combinatorial result used in many applications to show that a particular event happens with positive probability. It is known that if a large number of independent events each happen with positive probability, then there is positive (possibly exponentially small) probability that they all happen at the same time. The LLL, first proved by Erdős and Lovász [EL75], extends this result to the case of rare dependencies. Specifically, the LLL in its general form is the following [AS08]: Theorem 1. Let A 1, A 2,, A n be events in an arbitrary probability space. A directed graph G = (V, E) on the set of vertices V = {1, 2,, n} is called a dependency graph for the events A 1, A 2,, A n if for each i, 1 i n, the event A i is mutually independent of all the events {A j : (i, j) / E}. Suppose that G = (V, E) is a dependency digraph for the above events and suppose there are real numbers x 1,, x n such that 0 x i < 1 and Pr[A i ] x i (i,j) E (1 x j) for all 1 i n. Then [ n ] Pr A i i=1 n (1 x i ). i=1 1.1 The algorithmic perspective The LLL is an important tool used in the probabilistic method applied in randomized algorithms but also in a variety of approximation algorithms. It is in this latter context that we will discuss it. The main problem that arises in any endeavor to apply it algorithmically is its non-constructive aspect. Also, the LLL states that there is an assignment such that a particular event happens with positive probability. It does not, however, guarantee that the probability of it happening is not exponentially small in the dimensions of the problem. In this talk, we will describe the celebrated result of Moser and Tardos [MT10] that address these two problems and we will improve their result. This is joint work done with Bernhard Haeupler (MIT) and Barna Saha (UMD). For details, you can consult the original paper [HSS10]. First, let us present the usual framework in which the algorithmic versions of the LLL is used. Let {X 1, X 2,... X n } be independent random variables, and let A = {A 1, A 2,... A m } be a collection of m bad events, each determined by a subset of X j s. Let us also consider the ubiquitous version of the framework, with neighborhood relation Γ on A. 1
2 The main question arising from this framework is whether all A i can be avoided simultaneously. The problem becomes therefore, of finding an algorithm that outputs an assignment to all X j that guarantees all the bad events are avoided. Note that the output size is n. Also note that this framework is slightly more restrictive than the general framework in the sense that one can always transform the general setting into this framework by considering indicator variables for each of the bad events. 1.2 Results In this context, the main results of [HSS10] we are going to discuss are: A randomized, poly(n)-time algorithm that work for all known applications of the LLL, even when m poly(n), if we allow a tiny slack in the LLL condition. An application that algorithmically involves avoiding most of the bad events. Such a result is useful in MAX SAT-like problems (as opposed to SAT) in which there is a dichotomy between existing approaches. On one hand, in the case in which the dependency is small enough, LLL can be used and we can avoid all bad events. On the other hand, if the overlap is larger than the LLL threshold, the only known bound on the number of bad events is the trivial one given by computing the linearity of expectation on random assignments (in which each clause is violated with probability 2 k in MAX k-sat). 2 Symmetric LLL Before we begin discussing the main results, as a warm-up, we will present the symmetric version of the LLL. Suppose: max i Pr[A i ] p, and each A i has D neighbors in the dependency graph. Then, e p (D + 1) 1 implies Pr[ no A i holds ] > 0. Note: The typical case in which the symmetric version is applied is that in which D m, in which case the problem we encounter is that P r[ i A i] is inevitably small. Since the dependency graph is sparse, we can find a large independent set I of A i such that I m/(d +1). We therefore get that Pr[ i A i] Pr[ i I A i] = (1 p) m/(d+1) exp( mp/d). 2.1 Domatic partitions Now we are going to present an application of the symmetric LLL to the domatic number problem, as can be found in [FHKS02]. A set of vertices in a graph is a dominating set if every vertex outside the set has a neighbor in the set. The domatic number problem is that of partitioning the vertices of a graph into the maximum number of disjoint dominating sets. Let the domatic number 2
3 of a graph G be denote by D(G). As an interesting application of this problem to the problem of wireless coordination, please consult [CJBM02]. Another way of viewing the domatic number problem is that of coloring vertices with a maximum number of colors such that, for any vertex v, all colors are visible in N + (v)= inclusive neighborhood of vertex v. On the other hand, we can think of assignments in the LLL as being coloring of the vertices in the dependency graph. Using this approach, [FHKS02] provide an algorithm with a logarithmic approximation threshold, marking the rare incident of a maximization problem having such a threshold and no better. Before we begin describing the algorithm, however, let us remark that every graph G satisfies D(G) 1 and unless G contains an isolated node, D(G) 2 (consider a spanning tree in which the odd levels consist of one dominating set and the even levels of another one, therefore producing at least 2 dominating sets). At the same time, let (δ, ) = (min, max) degrees in G. Then we get that D(G) δ + 1 since a node of minimum degree must have some neighbor (or itself) in each of the disjoint dominating sets. 2.2 Domatic partitions assuming d-regularity Now we can proceed to present a restricted version of the algorithm in [FHKS02] that achieves a 3 ln d-approximation when G is d-regular. Independently color each vertex randomly with one of l d/(3 ln d) colors. For each (v, c) pair, define a Boolean variable A v,c to be true if there is no vertex of color c in N + (v), and to be false otherwise. Note that the events A v,c are bad events for us: if A v,c holds for some pair (v, c), then the coloring is not a domatic partition. For each pair (v, c), p = Pr[A v,c ] = (1 1/l) d+1 1/d 3. Now let us fix a A v,c an let us look at its dependence. We note that each event A v,c is independent of all events A w,c with dist(v, w) 3 since vertices of distance at least 3 from v are completely irrelevant for v. Since the number of vertices at distance 2 is < d and the number of possible colors c is l, we get that D < d 3 /(3 ln d). We also get that e p (D + 1) 1 and therefore, a good coloring exists. 2.3 Improving the Constant Improving the constant from 3 to 1 can be done by using an iterated approach of the LLL, a powerful methodology which has been successfully used in a variety of breakthrough results. For more information on this subject, we recommend [MR02]. The main difference from regular methods is that we apply a slow ( two-stage) partitioning that helps prune the dependencies in applying the LLL, leading to our improvement. The first partition is small-sized, but has properties much stronger than being domatic; the second partition refines the first, and crucially benefits from these useful properties of the first partition. 3
4 Specifically, we will color the vertices in two stages. Let χ 1 be the first coloring and χ 2 the second one. In the end, our coloring will be a pair of the two colors, one assigned by χ 1, and the other component assigned by χ 2. In the first stage, our goal is to color the vertices such that we can obtain a good range for the fraction of vertices in N + (u) that receives a particular color c, for any such color Note that this property is much stronger than the coloring being domatic, since being domatic just means that every color is visible in N + (u) and doesn t say anything about the fraction of neighbors that exhibit any color. By applying the LLL in the first stage, we ensure that we obtain such a good coloring. Then, in the second coloring, we can consider B u,c1,c 2 to be the bad event that there is no vertex of color (c 1, c 2 ) in N + (u). We will use the LLL to show that all these bad events can be avoided with positive probability. The main ingredient is bounding the number of dependencies of a fixed B u,c1,c 2. Let N u,c + = {v N + (u) : χ 1 (u) = c}. Then we get that B u,c1,c 2 simply says that all the elements of N + (u, c 1 ) got a χ 2 ( ) value different from c 2. Thus, we can check that B u,c1,c 2 only depends on the events in: S(u, c 1, c 2 ) = {B v,c 1,c 2 : v N + (N + u,c 1 ) and c 1 = c 1}. In the analysis, we get that S(u, c 1, c 2 ) is only 1+o(1) as compared to the previous dependence of 3+o(1), each of the constraints in the definition of S(u, c 1, c 2 ) saving us a factor of 1 o(1). 3 Asymmetric LLL Now let us get back to the general version of LLL and review its statement. x : A (0, 1) such that i : Pr[A i x(a i ) (1 x(a j )), A j Γ(A i ) We have that if then Pr[ i A i] i (1 x(a i)) > 0. The asymmetric version has numerous applications, some of them including: (hyper-)graph colorings and Ramsey numbers routing ([LMR94]) LP integrality gaps ([Fei08], [LLRS01]) edge-disjoint paths ([And10]). 3.1 Short proof of the general LLL We will review a short proof of the lemma, as it is given in [AS08], because it will give us insight into the main theorem of [FHKS02]. The idea of the proof is to split Pr[ i A i] into a product of conditional probabilities that we can prove a lower bound on. Notice that: 4
5 n Pr[ A i ] = (1 Pr[A 1 ]) (1 Pr[A 2 A 1 ]) (1 Pr[A n i=1 Hence, the first step is proving by induction on s, that for any S {1, n}, S = s < n and any i / S, Pr[A i j S A j ] x i. n 1 i=1 A i ]). The base case is trivial so now let s assume, for the inductive step, that the above holds for all s < s and prove that it also holds for s. Let S 1 = {j A : (i, j) E} and S 2 = S\S1. For clarity, we assume that G = (V, E) is the dependency graph. Then: P r[a i j S A j ] = P r[a i ( j S 1 A j ) l S 2 A l ] P r[ j S 1 A j. l S 2 A l ] Notice that since A i is mutually independent of the events in {A l : l S 2 }, we get that the numerator is upper bounded by Pr[A i A l ] = Pr[A i ] x i (1 x j ). l S 2 (i,j) E On the other hand, we can bound the denominator by the induction hypothesis. Let S 1 = j 1, j 2,, j r. If r = 0, the the denominator is 1 and our conclusion follows. Otherwise: Pr[A j1 A j2 A jr A l ] = (1 P r[a j1 A l ]) (1 P r[a j2 Aj 1 A l ]) l S 2 l S 2 l S 2 (1 P r[a jr A j1 A jr 1 (1 x j1 )(1 x j2 ) (1 x jr ) (1 x j ) (i,j) E l S 2 A l ]) Combining the two inequalities gives us the desired bound and completes the proof. 4 The Algorithmic Framework In trying to make the proof constructive, it is natural to consider the following trivial algorithm: repeat pick random assignments for all X j until no A i holds The LLL theorem tells us that is the LLL conditions holds, then the algorithm finds a satisfying assignment with positive probability. However, the running time might be exponential in m, let alone n. 5
6 4.1 The Moser-Tardos breakthrough Given such a situation, several algorithmic versions of the LLL have emerged, ([Bec91], [Alo91], [MR98], [CS00], [Sri08], [Mos09]) culminating in the Moser-Tardos(MT) breakthrough [MT10]: start with an arbitrary assignment while event A i that holds do assign new random values to the variables of A i The analysis of the algorithm is based on the following theorem from [MT10]: Theorem 2. If the LLL-conditions hold, then the above algorithm finds a satisfying assignment within an expected x(a i ) i 1 x(a i ) iterations. Proof. In order to prove this, [MT10] describe the process of obtaining such an assignment as a branching process that corresponds to decisions of resampling some variables. The log of the execution is the list of events in A as they have been selected for resampling at each step. They associate with each resampling step a witness tree that can serve as a justification for the necessity of the correction step. A particular witness tree occurs in the log of the execution if there exists a step t such that the witness tree is rooted at it. The idea of the proof is to upper bound the number of resamplings each event A A needs (remember that the process finishes when no such resampling is needed). The authors do that in two steps: first, they upper bound the probability that a fixed witness tree occurs in a random log produced by the algorithm second, they obtain a probability distribution over different witness trees. For each such witness tree, they consider a Galton-Watson branching process that attempts to construct it. They start with the singleton event A as the root and then, in each round, they add to the vertex v a child for the event B (v or adjacent to v in the dependency graph) with probability x(b) or skip it with probability 1 x(b). With these two ingredients, the authors are able to upper bound the expectation of the number of occurrences of A in the log simply by summing the bounds for a fixed witness tree rooted at A on the probabilities of the occurrences of the different witness trees. The bound they obtain x(a) for a particular A is 1 x(a). Summing over all such A therefore gives us the upper bound on the expected number of total resampling. 4.2 LLL-distribution and the MT- algorithm One observation that needs to be made is that the MT algorithm doesn t just provide an assignment in which all bad events are avoided. In fact, it provides a randomly chosen assignment from the distribution that is obtained when one conditions on no bad events holding. [HSS10] uses this observation to introduce a new proof-concept based on the (conditional) LLL-distribution D. Some very useful properties are known for D: 6
7 Theorem 3. Pr D [B] = Pr[B i A i ] Pr(B) Aj Γ(B)(1 x(a j )) 1 Proof. Looking at the proof for the general LLL, one notices that it can the induction can be applied to any event B conditioned on a the bad events not happening, where B is determined by X i s. Consider any S {1,, n}, S = s < n, then let S 1 = S Γ(B) and S 2 = S\S 1. Then we get as before that: Pr[B j S A j ] = Pr[B ( j S 1 A j ) l S 2 A l ] Pr[ j S 1 A j. l S 2 A l ] Following the calculations just as before, we get that the numerator is smaller than Pr[B] and the denominator is bigger than A j Γ(B) (1 x(a j)). Informally, if B depends not too heavily on the events A j, then the probability places on B by D is not much more than the unconditional probability Pr(B). Such bounds in combination with further probabilistic analysis can be used to give interesting(nonconstructive) results. The main result of [HSS10] is that the MT algorithm has an output distribution that approximates the LLL-distribution D: in that for every B, the same upper bound as above holds as well. This can be used to make the probabilistic proofs that use the LLL-condition constructive. In particular, we can modify the analysis of the MT algorithm to upper bound the probability that an event B was true at least once during the execution of the MT algorithm. Note that B does not have to be in A. The witness trees that certify the first time B become true are the ones that have B as a root and all non-root nodes from A\B. Just as in the MT analysis, we calculate the probability that the Galton-Watson process builds a witness tree rooted at B and then calculate expected number of these witness trees via a union bound. The upper bound that we get is exactly Pr[B] ( (1 x(a j ))) 1, where the term Pr[B] accounts for the fact that the root-event B A j Γ(B) has to be true as well. Such insight will prove significant in developing a more refined version of the MT algorithm, as it is described in the following sections. 4.3 LLL- Applications with exponentially many events For now, however, let us backtrack and look at the greater picture. The main question that arises in the algorithmic aspect of the LLL is the situation in which m is exponentially large. Such situations can be encountered in a variety of problems such as: Acyclic edge coloring Non-repetitive coloring Santa Claus problem 7
8 Edge-disjoint paths In fact, when running MT, we encounter the following problems: 1. the expected number of resamplings is i 2. we need to represent the bad events x(a i ) 1 x(a i ) 3. we need to verify a solution/ find some A i that currently holds The first problem can be solved by proving an upper bound: Let δ = min i Pr[A i ]. Then, x(a i ) 1 x(a i ) E[ # iterations of MT] i ( 1 x(a i )) max i 1 x(a i i ) 1 O(n log(1/δ)) max i 1 x(a i ) In all applications known to us, log( 1 δ ) O(n log n). While we have good bounds on the running time of MT even for application with Ω(n) m poly(n) many events, it unfortunately often fails to be directly applicable when m becomes superpolynomial in n. The reason is that maintaining bad events implicitly and running the resampling process requires an efficient way to find violated events. In many examples with super-polynomially many events, finding violated events or even just verifying a good assignment is hard. The solution that [HSS10] give to this problem is: find a core-subset of the events, of polynomial size, and run MT on that bound the probabilities of non-core events using Theorem 3. use a union bound over these probabilities to prove that with high probability all of the A i are avoided Essentially what this algorithm does is it uses the randomness in the output distribution of the MT-algorithm run on the smaller core. Here is where the theorem about the LLL-distribution becomes helpful. The non-core events will have small probabilities and will be sparsely connected to core events and as such their probabilities in the LLL-distribution and therefore also the output distribution of the algorithm does not blow up by much. As another way to look at the proof, one can imagine redefining the witness trees to have only core events as non-root nodes, just like we did in the proof of Theorem 3. Then, with this modified Galton-Watson process, we can grow witness trees from a log starting with a root event that holds at a certain point in time. This guarantees that we capture non-core events happening even though they are never resampled (since we never check whether such events hold or not). The following theorem summarizes the idea: 8
9 Theorem 4. If ɛ (0, 1) such that for all A i, then: Pr[A] 1 ɛ x(a i ) A j Γ(A i ) for any p 1 poly(n), {A i : Pr[A i ] p} poly(n); (1 x(a j )), If log 1 δ poly(n) and the above core is checkable, then for any desired constant c > 0, a Monte Carlo algorithm(with p n c/ɛ ) that terminates within O( n ɛ log n ɛ ) resamplings and returns a good assignment with probability at least 1 n c. The idea is that, for every p 1 poly(n), the set {A i : Pr[A i ] p} has size at most poly(n), and is thus essentially always an efficiently verifiable core subset of A. In order to deal with the non-core events, we can use Theorem 3, in the sense that the underlying dependency graph G is very dense (the m vertices of G can be partitioned into n cliques according to the variables the bad events depend on) and the nature of the LLL-conditions force highly connected events to have small probabilities and x-values. The ɛ-slack provides an extra multiplicative factor over the LLL-conditions but by choosing p = n O(1/ɛ), we can make the union bound small, at most n c. 4.4 Algorithmic results This approach has led to a number of efficient algorithms for: O(1)-approximation for the Santa Claus problem(feige s proof [AFS08] made constructive) non-repetitive coloring(proof of Alon-Grytczuk-Hauszczak-Riordan [AGHR02] made constructive ) acyclic edge coloring edge-disjoint paths([and10]) By the Santa Claus problem we mean the restricted assignment version of the max-min allocation problem for indivisible goods. In the max-min allocation problem, there is a set of n items and m platers. The value(utility) of item j to player i is p i,j 0. An item can be assigned to one player and the total valuation of the items received by a player is the sum of the individual valuations of each item, p(i, j), for player i and item j. The goal is to maximize the minimum total valuation of the items received by any player, that is, to maximize min i j S i p(i, j), where S i is the subset of items player i receives. In the restricted version of the max-min allocation problem, each item has an intrinsic value, for every player i, p i,j is either p j or 0. [BS06] considered an LP relaxation to the problem known as the configuration LP. [AFS08] showed that the LP has a constant integrality gap but the proof cannot be made constructive because it is heavily based on repeated reductions that apply the asymmetric LLL to exponentially many events. In this context, Theorem 4 can be easily and directly applied to constructivize the proof. 9
10 5 Allowing Some A i to Hold As mentioned before, another result refers to interpolating between the LLL and the linearity of expectation. [HSS10] obtain the following result: Theorem 5. In the symmetric LLL with p and D, if D α (1/(ep) 1) where 1 < α < e, then we can make at most (e ln(α)/α) mp of the A i to hold, in randomized poly(m) time. The main observation is that, when α = 1, we have the standard LLL and all bad events can be avoided. If α > e, then the best one can do is linearity of expectation. In the case in which 1 < α < e, the core is chosen at random but the number of non-core events is not too big so that linearity of expectation can be applied. As mentioned before, allowing most bad events to be avoided can be very useful in MAX k-sat problems. 6 A Different Framework [KS11] introduce a different framework of looking at LLL, based on the work of Shearer ([She85]). Shearer has found an exact characterization of the sequence of probabilities for which all bad events can be avoided. The difference is that the dependency graph in the MT framework(variable, algorithmic version) is more restrictive than the dependency graph based on probabilistic independence (general version). [KS11] make the link by relating Shearer s framework (coming from the general version) with the analysis of the MT algorithm (coming from the variable version). The main technique is to construct Shearer condition equivalents that correspond to steps in the analysis of the MT algorithm. For example, instead of the Galton-Watson branching processes, they provide bounds using the spectral norm of the Stable Set Matrix. They obtain results on both sides. On one hand, they prove that the sequential and parallel versions of the MT algorithm are efficient up to Shearer s bound and extend our result by showing that the running time is not only polynomial in the number of events but also in the number of variables. In the opposite direction, they reprove Shearer s result for the general LLL, making a tighter connection between the general and variable versions of LLL. 7 Conclusion and Open Problems As we have seen, the LLL is an elegant statement that provides successful results in randomized algorithms and surprising results in approximation algorithms. The main problem we have encountered is the difficulty of making it constructive and efficient in the context of algorithms. The MT framework is a celebrated result in that direction. In addition, various problems inspire further refined approaches, such as considering a polynomially sized core and allowing some bad events to hold. In the end, we would like to leave the reader with a few open questions that highlight potential directions in which the application of LLL can be taken: 10
11 Is the term e ln(α)/α tight? Can the algorithm be derandomized? Can we further analyze the dependencies among non-core events, such that we use something else instead of a union-bound? How much slack is really needed? Can we use [KS11] as a possible guideline for generalizing MT? Our approach doesn t work in giving results about the Lopsided Local Lemma (a variant of the LLL that also distinguishes between negative and positive correlation [AS08]). However, the framework used by [KS11] is more abstract than the MT framework which heavily relies on the underlying assignment. Can a full understanding of the setting extend some results? References [AFS08] Arash Asadpour, Uriel Feige, and Amin Saberi. Santa Claus Meets Hypergraph Matchings. In APPROX-RANDOM, pages 10 20, [AGHR02] Noga Alon, Jaroslaw Grytczuk, Mariusz Haluszczak, and Oliver Riordan. Nonrepetitive colorings of graphs. Random Struct. Algorithms, 21(3-4): , [Alo91] Noga Alon. A Parallel Algorithmic Version of the Local Lemma. Random Struct. Algorithms, 2(4): , [And10] Matthew Andrews. Approximation Algorithms for the Edge-Disjoint Paths Problem via Raecke Decompositions. In FOCS, pages , [AS08] Noga Alon and Joel H. Spencer. The Probabilistic Method. Wiley, New York, [Bec91] [BS06] József Beck. An Algorithmic Approach to the Lovász Local Lemma. i. Random Struct. Algorithms, 2(4): , Nikhil Bansal and Maxim Sviridenko. The Santa Claus problem. In Proceedings of the thirty-eighth annual ACM symposium on Theory of computing, STOC 06, pages 31 40, New York, NY, USA, ACM. [CJBM02] Benjie Chen, Kyle Jamieson, Hari Balakrishnan, and Robert Morris. Span: An Energy- Efficient Coordination Algorithm for Topology Maintenance in Ad Hoc Wireless Networks. Wireless Networks, pages , [CS00] Artur Czumaj and Christian Scheideler. Coloring non-uniform hypergraphs: a new algorithmic approach to the general Lovász local lemma. In SODA, pages 30 39, [EL75] Paul Erdős and László Lovász. Problems and results on 3-chromatic hypergraphs and some related questions. In R. Rado A. Hajnal and V.T. Sós, editors, Infinite and Finite Sets, pages North-Holland, Amsterdam, [Fei08] Uriel Feige. On allocations that maximize fairness. In SODA, pages ,
12 [FHKS02] [HSS10] [KS11] [LLRS01] Uriel Feige, Magnús M. Halldórsson, Guy Kortsarz, and Aravind Srinivasan. Approximating the Domatic Number. SIAM J. Comput., 32(1): , Bernhard Haeupler, Barna Saha, and Aravind Srinivasan. New Constructive Aspects of the Lovasz Local Lemma. In Proceedings of the 2010 IEEE 51st Annual Symposium on Foundations of Computer Science, FOCS 10, pages , Washington, DC, USA, IEEE Computer Society. Kashyap Babu Rao Kolipaka and Mario Szegedy. Moser and Tardos meet Lovász. In STOC, pages , Frank Thomson Leighton, Chi-Jen Lu, Satish Rao, and Aravind Srinivasan. New Algorithmic Aspects of the Local Lemma with Applications to Routing and Partitioning. SIAM J. Comput., 31(2): , [LMR94] F. T. Leighton, Bruce M. Maggs, and Satish B. Rao. Packet routing and jobshop scheduling in o(congestion+dilation) steps. Combinatorica, 14: , /BF [Mos09] [MR98] Robin A. Moser. A constructive proof of the Lovász local lemma. In STOC, pages , Michael Molloy and Bruce A. Reed. Further Algorithmic Aspects of the Local Lemma. In STOC, pages , [MR02] M. Molloy and B. Reed. Graph colouring and the probabilistic method [MT10] Robin A. Moser and Gábor Tardos. A constructive proof of the general lovász local lemma. J. ACM, 57(2), [She85] J. Shearer. On a problem of spencer. Combinatorica, 5: , /BF [Sri08] Aravind Srinivasan. Improved algorithmic versions of the Lovász Local Lemma. In SODA, pages ,
An Algorithmic Proof of the Lopsided Lovász Local Lemma (simplified and condensed into lecture notes)
An Algorithmic Proof of the Lopsided Lovász Local Lemma (simplified and condensed into lecture notes) Nicholas J. A. Harvey University of British Columbia Vancouver, Canada nickhar@cs.ubc.ca Jan Vondrák
More informationThe Lovász Local Lemma : A constructive proof
The Lovász Local Lemma : A constructive proof Andrew Li 19 May 2016 Abstract The Lovász Local Lemma is a tool used to non-constructively prove existence of combinatorial objects meeting a certain conditions.
More informationAn extension of the Moser-Tardos algorithmic local lemma
An extension of the Moser-Tardos algorithmic local lemma Wesley Pegden January 26, 2013 Abstract A recent theorem of Bissacot, et al. proved using results about the cluster expansion in statistical mechanics
More informationLecture 10 Algorithmic version of the local lemma
Lecture 10 Algorithmic version of the local lemma Uriel Feige Department of Computer Science and Applied Mathematics The Weizman Institute Rehovot 76100, Israel uriel.feige@weizmann.ac.il June 9, 2014
More informationLecture 24: April 12
CS271 Randomness & Computation Spring 2018 Instructor: Alistair Sinclair Lecture 24: April 12 Disclaimer: These notes have not been subjected to the usual scrutiny accorded to formal publications. They
More informationarxiv: v5 [cs.ds] 2 Oct 2011
New Constructive Aspects of the Lovász Local Lemma Bernhard Haeupler Barna Saha Aravind Srinivasan October 4, 2011 arxiv:1001.1231v5 [cs.ds] 2 Oct 2011 Abstract The Lovász Local Lemma (LLL) is a powerful
More informationGlobal Optima from Local Algorithms
Global Optima from Local Algorithms Dimitris Achlioptas 1 Fotis Iliopoulos 2 1 UC Santa Cruz 2 UC Berkeley Dimitris Achlioptas (UC Santa Cruz) Global Optima from Local Algorithms IMA Graphical Models 2015
More informationComputing the Independence Polynomial: from the Tree Threshold Down to the Roots
1 / 16 Computing the Independence Polynomial: from the Tree Threshold Down to the Roots Nick Harvey 1 Piyush Srivastava 2 Jan Vondrák 3 1 UBC 2 Tata Institute 3 Stanford SODA 2018 The Lovász Local Lemma
More informationCollege of Computer & Information Science Fall 2009 Northeastern University 22 September 2009
College of Computer & Information Science Fall 2009 Northeastern University 22 September 2009 CS 7880: Algorithmic Power Tools Scribe: Eric Miles Lecture Outline: General Lovász Local Lemma Two Simple
More informationThe Lovász Local Lemma: constructive aspects, stronger variants and the hard core model
The Lovász Local Lemma: constructive aspects, stronger variants and the hard core model Jan Vondrák 1 1 Dept. of Mathematics Stanford University joint work with Nick Harvey (UBC) The Lovász Local Lemma
More informationImproved Bounds for Flow Shop Scheduling
Improved Bounds for Flow Shop Scheduling Monaldo Mastrolilli and Ola Svensson IDSIA - Switzerland. {monaldo,ola}@idsia.ch Abstract. We resolve an open question raised by Feige & Scheideler by showing that
More informationCS6999 Probabilistic Methods in Integer Programming Randomized Rounding Andrew D. Smith April 2003
CS6999 Probabilistic Methods in Integer Programming Randomized Rounding April 2003 Overview 2 Background Randomized Rounding Handling Feasibility Derandomization Advanced Techniques Integer Programming
More informationarxiv: v1 [math.co] 18 Nov 2017
Short proofs for generalizations of the Lovász Local Lemma: Shearer s condition and cluster expansion arxiv:171106797v1 [mathco] 18 Nov 2017 Nicholas J A Harvey Abstract Jan Vondrák The Lovász Local Lemma
More informationAn asymptotically tight bound on the adaptable chromatic number
An asymptotically tight bound on the adaptable chromatic number Michael Molloy and Giovanna Thron University of Toronto Department of Computer Science 0 King s College Road Toronto, ON, Canada, M5S 3G
More informationDeterministic Algorithms for the Lovász Local Lemma
Deterministic Algorithms for the Lovász Local Lemma The MIT Faculty has made this article openly available. Please share how this access benefits you. Your story matters. Citation As Published Publisher
More informationProbabilistic Proofs of Existence of Rare Events. Noga Alon
Probabilistic Proofs of Existence of Rare Events Noga Alon Department of Mathematics Sackler Faculty of Exact Sciences Tel Aviv University Ramat-Aviv, Tel Aviv 69978 ISRAEL 1. The Local Lemma In a typical
More informationPCPs and Inapproximability Gap-producing and Gap-Preserving Reductions. My T. Thai
PCPs and Inapproximability Gap-producing and Gap-Preserving Reductions My T. Thai 1 1 Hardness of Approximation Consider a maximization problem Π such as MAX-E3SAT. To show that it is NP-hard to approximation
More informationThe Lefthanded Local Lemma characterizes chordal dependency graphs
The Lefthanded Local Lemma characterizes chordal dependency graphs Wesley Pegden March 30, 2012 Abstract Shearer gave a general theorem characterizing the family L of dependency graphs labeled with probabilities
More informationOn (1, ε)-restricted Assignment Makespan Minimization
On (1, ε)-restricted Assignment Makespan Minimization Deeparnab Chakrabarty Sanjeev Khanna Shi Li Abstract Makespan minimization on unrelated machines is a classic problem in approximation algorithms.
More informationSanta Claus Schedules Jobs on Unrelated Machines
Santa Claus Schedules Jobs on Unrelated Machines Ola Svensson (osven@kth.se) Royal Institute of Technology - KTH Stockholm, Sweden March 22, 2011 arxiv:1011.1168v2 [cs.ds] 21 Mar 2011 Abstract One of the
More informationLecture Notes CS:5360 Randomized Algorithms Lecture 20 and 21: Nov 6th and 8th, 2018 Scribe: Qianhang Sun
1 Probabilistic Method Lecture Notes CS:5360 Randomized Algorithms Lecture 20 and 21: Nov 6th and 8th, 2018 Scribe: Qianhang Sun Turning the MaxCut proof into an algorithm. { Las Vegas Algorithm Algorithm
More informationABSTRACT THE LOVÁSZ LOCAL LEMMA. Dissertation directed by: Professor Aravind Srinivasan, Department of Computer Science
ABSTRACT Title of Dissertation ALGORITHMS AND GENERALIZATIONS FOR THE LOVÁSZ LOCAL LEMMA David G. Harris, Doctor of Philosophy, 2015 Dissertation directed by: Professor Aravind Srinivasan, Department of
More informationU.C. Berkeley CS294: Beyond Worst-Case Analysis Handout 3 Luca Trevisan August 31, 2017
U.C. Berkeley CS294: Beyond Worst-Case Analysis Handout 3 Luca Trevisan August 3, 207 Scribed by Keyhan Vakil Lecture 3 In which we complete the study of Independent Set and Max Cut in G n,p random graphs.
More informationApplications of the Lopsided Lovász Local Lemma Regarding Hypergraphs
Regarding Hypergraphs Ph.D. Dissertation Defense April 15, 2013 Overview The Local Lemmata 2-Coloring Hypergraphs with the Original Local Lemma Counting Derangements with the Lopsided Local Lemma Lopsided
More informationU.C. Berkeley CS294: Beyond Worst-Case Analysis Handout 8 Luca Trevisan September 19, 2017
U.C. Berkeley CS294: Beyond Worst-Case Analysis Handout 8 Luca Trevisan September 19, 2017 Scribed by Luowen Qian Lecture 8 In which we use spectral techniques to find certificates of unsatisfiability
More informationLecture 4: Inequalities and Asymptotic Estimates
CSE 713: Random Graphs and Applications SUNY at Buffalo, Fall 003 Lecturer: Hung Q. Ngo Scribe: Hung Q. Ngo Lecture 4: Inequalities and Asymptotic Estimates We draw materials from [, 5, 8 10, 17, 18].
More informationLecture Constructive Algorithms for Discrepancy Minimization
Approximation Algorithms Workshop June 13, 2011, Princeton Lecture Constructive Algorithms for Discrepancy Minimization Nikhil Bansal Scribe: Daniel Dadush 1 Combinatorial Discrepancy Given a universe
More informationarxiv: v1 [cs.ds] 8 Dec 2016
A CONSTRUCTIVE ALGORITHM FOR THE LOVÁSZ LOCAL LEMMA ON PERMUTATIONS 1 DAVID G. HARRIS 2 AND ARAVIND SRINIVASAN 3 arxiv:1612.02663v1 [cs.ds] 8 Dec 2016 Abstract. While there has been significant progress
More informationLecture 5. Shearer s Lemma
Stanford University Spring 2016 Math 233: Non-constructive methods in combinatorics Instructor: Jan Vondrák Lecture date: April 6, 2016 Scribe: László Miklós Lovász Lecture 5. Shearer s Lemma 5.1 Introduction
More informationA Different Perspective For Approximating Max Set Packing
Weizmann Institute of Science Thesis for the degree Master of Science Submitted to the Scientific Council of the Weizmann Institute of Science Rehovot, Israel A Different Perspective For Approximating
More informationOn (1, ε)-restricted Assignment Makespan Minimization
On 1, ε)-restricted Assignment Makespan Minimization Deeparnab Chakrabarty Sanjeev Khanna Shi Li Abstract Makespan minimization on unrelated machines is a classic problem in approximation algorithms. No
More informationn Boolean variables: x 1, x 2,...,x n 2 {true,false} conjunctive normal form:
Advanced Algorithms k-sat n Boolean variables: x 1, x 2,...,x n 2 {true,false} conjunctive normal form: k-cnf = C 1 ^C 2 ^ ^C m Is φ satisfiable? m clauses: C 1,C 2,...,C m each clause C i = `i1 _ `i2
More informationAdvanced Algorithms 南京大学 尹一通
Advanced Algorithms 南京大学 尹一通 Constraint Satisfaction Problem variables: (CSP) X = {x 1, x2,..., xn} each variable ranges over a finite domain Ω an assignment σ ΩX assigns each variable a value in Ω constraints:
More informationThe Lopsided Lovász Local Lemma
Department of Mathematics Nebraska Wesleyan University With Linyuan Lu and László Székely, University of South Carolina Note on Probability Spaces For this talk, every a probability space Ω is assumed
More informationOn the Beck-Fiala Conjecture for Random Set Systems
On the Beck-Fiala Conjecture for Random Set Systems Esther Ezra Shachar Lovett arxiv:1511.00583v1 [math.co] 2 Nov 2015 November 3, 2015 Abstract Motivated by the Beck-Fiala conjecture, we study discrepancy
More informationCMPUT 675: Approximation Algorithms Fall 2014
CMPUT 675: Approximation Algorithms Fall 204 Lecture 25 (Nov 3 & 5): Group Steiner Tree Lecturer: Zachary Friggstad Scribe: Zachary Friggstad 25. Group Steiner Tree In this problem, we are given a graph
More informationOut-colourings of Digraphs
Out-colourings of Digraphs N. Alon J. Bang-Jensen S. Bessy July 13, 2017 Abstract We study vertex colourings of digraphs so that no out-neighbourhood is monochromatic and call such a colouring an out-colouring.
More informationColoring Uniform Hypergraphs with Few Colors*
Coloring Uniform Hypergraphs with Few Colors* Alexandr Kostochka 1,2 1 University of Illinois at Urbana-Champaign, Urbana, Illinois 61801; e-mail: kostochk@mathuiucedu 2 Institute of Mathematics, Novosibirsk
More informationLecture 1 : Probabilistic Method
IITM-CS6845: Theory Jan 04, 01 Lecturer: N.S.Narayanaswamy Lecture 1 : Probabilistic Method Scribe: R.Krithika The probabilistic method is a technique to deal with combinatorial problems by introducing
More informationOn Allocating Goods to Maximize Fairness
On Allocating Goods to Maximize Fairness Deeparnab Chakrabarty Julia Chuzhoy Sanjeev Khanna February 24, 2012 Abstract We consider the Max-Min Allocation problem: given a set A of m agents and a set I
More informationAn estimate for the probability of dependent events
Statistics and Probability Letters 78 (2008) 2839 2843 Contents lists available at ScienceDirect Statistics and Probability Letters journal homepage: www.elsevier.com/locate/stapro An estimate for the
More informationLecture 5: January 30
CS71 Randomness & Computation Spring 018 Instructor: Alistair Sinclair Lecture 5: January 30 Disclaimer: These notes have not been subjected to the usual scrutiny accorded to formal publications. They
More informationTwo-coloring random hypergraphs
Two-coloring random hypergraphs Dimitris Achlioptas Jeong Han Kim Michael Krivelevich Prasad Tetali December 17, 1999 Technical Report MSR-TR-99-99 Microsoft Research Microsoft Corporation One Microsoft
More informationOn shredders and vertex connectivity augmentation
On shredders and vertex connectivity augmentation Gilad Liberman The Open University of Israel giladliberman@gmail.com Zeev Nutov The Open University of Israel nutov@openu.ac.il Abstract We consider the
More informationA CONSTRUCTIVE ALGORITHM FOR THE LOVÁSZ LOCAL LEMMA ON PERMUTATIONS 1
A CONSTRUCTIVE ALGORITHM FOR THE LOVÁSZ LOCAL LEMMA ON PERMUTATIONS 1 DAVID G. HARRIS 2 AND ARAVIND SRINIVASAN 3 Abstract. While there has been significant progress on algorithmic aspects of the Lovász
More informationSize and degree anti-ramsey numbers
Size and degree anti-ramsey numbers Noga Alon Abstract A copy of a graph H in an edge colored graph G is called rainbow if all edges of H have distinct colors. The size anti-ramsey number of H, denoted
More informationZ Algorithmic Superpower Randomization October 15th, Lecture 12
15.859-Z Algorithmic Superpower Randomization October 15th, 014 Lecture 1 Lecturer: Bernhard Haeupler Scribe: Goran Žužić Today s lecture is about finding sparse solutions to linear systems. The problem
More informationIncreasing the Span of Stars
Increasing the Span of Stars Ning Chen Roee Engelberg C. Thach Nguyen Prasad Raghavendra Atri Rudra Gynanit Singh Department of Computer Science and Engineering, University of Washington, Seattle, WA.
More informationAcyclic subgraphs with high chromatic number
Acyclic subgraphs with high chromatic number Safwat Nassar Raphael Yuster Abstract For an oriented graph G, let f(g) denote the maximum chromatic number of an acyclic subgraph of G. Let f(n) be the smallest
More informationThe Probabilistic Method
The Probabilistic Method Janabel Xia and Tejas Gopalakrishna MIT PRIMES Reading Group, mentors Gwen McKinley and Jake Wellens December 7th, 2018 Janabel Xia and Tejas Gopalakrishna Probabilistic Method
More informationProperly colored Hamilton cycles in edge colored complete graphs
Properly colored Hamilton cycles in edge colored complete graphs N. Alon G. Gutin Dedicated to the memory of Paul Erdős Abstract It is shown that for every ɛ > 0 and n > n 0 (ɛ), any complete graph K on
More informationLecture # 12. Agenda: I. Vector Program Relaxation for Max Cut problem. Consider the vectors are in a sphere. Lecturer: Prof.
Lecture # 12 Lecturer: Prof. Allan Borodin Scribe: Yeleiny Bonilla Agenda: I. Finish discussion of Vector Program Relaxation for Max-Cut problem. II. Briefly discuss same approach for Max-2-Sat. III. The
More informationThe Budgeted Unique Coverage Problem and Color-Coding (Extended Abstract)
The Budgeted Unique Coverage Problem and Color-Coding (Extended Abstract) Neeldhara Misra 1, Venkatesh Raman 1, Saket Saurabh 2, and Somnath Sikdar 1 1 The Institute of Mathematical Sciences, Chennai,
More informationNotes for Lecture 2. Statement of the PCP Theorem and Constraint Satisfaction
U.C. Berkeley Handout N2 CS294: PCP and Hardness of Approximation January 23, 2006 Professor Luca Trevisan Scribe: Luca Trevisan Notes for Lecture 2 These notes are based on my survey paper [5]. L.T. Statement
More informationPacking and Covering Dense Graphs
Packing and Covering Dense Graphs Noga Alon Yair Caro Raphael Yuster Abstract Let d be a positive integer. A graph G is called d-divisible if d divides the degree of each vertex of G. G is called nowhere
More informationRobin Moser makes Lovász Local Lemma Algorithmic! Notes of Joel Spencer
Robin Moser makes Lovász Local Lemma Algorithmic! Notes of Joel Spencer 1 Preliminaries The idea in these notes is to explain a new approach of Robin Moser 1 to give an algorithm for the Lovász Local Lemma.
More informationOn a hypergraph matching problem
On a hypergraph matching problem Noga Alon Raphael Yuster Abstract Let H = (V, E) be an r-uniform hypergraph and let F 2 V. A matching M of H is (α, F)- perfect if for each F F, at least α F vertices of
More informationApproximability of Packing Disjoint Cycles
Approximability of Packing Disjoint Cycles Zachary Friggstad Mohammad R. Salavatipour Department of Computing Science University of Alberta Edmonton, Alberta T6G 2E8, Canada zacharyf,mreza@cs.ualberta.ca
More informationEdge-disjoint induced subgraphs with given minimum degree
Edge-disjoint induced subgraphs with given minimum degree Raphael Yuster Department of Mathematics University of Haifa Haifa 31905, Israel raphy@math.haifa.ac.il Submitted: Nov 9, 01; Accepted: Feb 5,
More informationAditya Bhaskara CS 5968/6968, Lecture 1: Introduction and Review 12 January 2016
Lecture 1: Introduction and Review We begin with a short introduction to the course, and logistics. We then survey some basics about approximation algorithms and probability. We also introduce some of
More informationCommunication Complexity of Combinatorial Auctions with Submodular Valuations
Communication Complexity of Combinatorial Auctions with Submodular Valuations Shahar Dobzinski Weizmann Institute of Science Rehovot, Israel dobzin@gmail.com Jan Vondrák IBM Almaden Research Center San
More informationLecture notes on OPP algorithms [Preliminary Draft]
Lecture notes on OPP algorithms [Preliminary Draft] Jesper Nederlof June 13, 2016 These lecture notes were quickly assembled and probably contain many errors. Use at your own risk! Moreover, especially
More informationThe concentration of the chromatic number of random graphs
The concentration of the chromatic number of random graphs Noga Alon Michael Krivelevich Abstract We prove that for every constant δ > 0 the chromatic number of the random graph G(n, p) with p = n 1/2
More informationProbabilistic Constructions of Computable Objects and a Computable Version of Lovász Local Lemma.
Probabilistic Constructions of Computable Objects and a Computable Version of Lovász Local Lemma. Andrei Rumyantsev, Alexander Shen * January 17, 2013 Abstract A nonconstructive proof can be used to prove
More informationStability of the path-path Ramsey number
Worcester Polytechnic Institute Digital WPI Computer Science Faculty Publications Department of Computer Science 9-12-2008 Stability of the path-path Ramsey number András Gyárfás Computer and Automation
More informationFPT hardness for Clique and Set Cover with super exponential time in k
FPT hardness for Clique and Set Cover with super exponential time in k Mohammad T. Hajiaghayi Rohit Khandekar Guy Kortsarz Abstract We give FPT-hardness for setcover and clique with super exponential time
More informationA constructive algorithm for the Lovász Local Lemma on permutations
A constructive algorithm for the Lovász Local Lemma on permutations David G. Harris Aravind Srinivasan Abstract While there has been significant progress on algorithmic aspects of the Lovász Local Lemma
More informationA simple LP relaxation for the Asymmetric Traveling Salesman Problem
A simple LP relaxation for the Asymmetric Traveling Salesman Problem Thành Nguyen Cornell University, Center for Applies Mathematics 657 Rhodes Hall, Ithaca, NY, 14853,USA thanh@cs.cornell.edu Abstract.
More informationNP Completeness and Approximation Algorithms
Chapter 10 NP Completeness and Approximation Algorithms Let C() be a class of problems defined by some property. We are interested in characterizing the hardest problems in the class, so that if we can
More informationThe Lopsided Lovász Local Lemma
Joint work with Linyuan Lu and László Székely Georgia Southern University April 27, 2013 The lopsided Lovász local lemma can establish the existence of objects satisfying several weakly correlated conditions
More informationLecture 5: Probabilistic tools and Applications II
T-79.7003: Graphs and Networks Fall 2013 Lecture 5: Probabilistic tools and Applications II Lecturer: Charalampos E. Tsourakakis Oct. 11, 2013 5.1 Overview In the first part of today s lecture we will
More informationTree Decomposition of Graphs
Tree Decomposition of Graphs Raphael Yuster Department of Mathematics University of Haifa-ORANIM Tivon 36006, Israel. e-mail: raphy@math.tau.ac.il Abstract Let H be a tree on h 2 vertices. It is shown
More informationLecture 11: Generalized Lovász Local Lemma. Lovász Local Lemma
Lecture 11: Generalized Recall We design an experiment with independent random variables X 1,..., X m We define bad events A 1,..., A n where) the bad event A i depends on the variables (X k1,..., X kni
More informationLovász Local Lemma: A Survey of Constructive and Algorithmic Aspects, with an Application to Packet Routing
Lovász Local Lemma: A Survey of Constructive and Algorithmic Aspects, with an Application to Packet Routing CPSC 536N - Term Porject Bader N. Alahmad 1 Introduction The Lovász Local Lemma (LLL) is a mathematical
More informationRainbow Hamilton cycles in uniform hypergraphs
Rainbow Hamilton cycles in uniform hypergraphs Andrzej Dude Department of Mathematics Western Michigan University Kalamazoo, MI andrzej.dude@wmich.edu Alan Frieze Department of Mathematical Sciences Carnegie
More informationA Linear Round Lower Bound for Lovasz-Schrijver SDP Relaxations of Vertex Cover
A Linear Round Lower Bound for Lovasz-Schrijver SDP Relaxations of Vertex Cover Grant Schoenebeck Luca Trevisan Madhur Tulsiani Abstract We study semidefinite programming relaxations of Vertex Cover arising
More informationMatroids and Integrality Gaps for Hypergraphic Steiner Tree Relaxations
Matroids and Integrality Gaps for Hypergraphic Steiner Tree Relaxations Rico Zenklusen Johns Hopkins University Joint work with: Michel Goemans, Neil Olver, Thomas Rothvoß 2 / 16 The Steiner Tree problem
More informationInduced subgraphs of prescribed size
Induced subgraphs of prescribed size Noga Alon Michael Krivelevich Benny Sudakov Abstract A subgraph of a graph G is called trivial if it is either a clique or an independent set. Let q(g denote the maximum
More informationOn the Algorithmic Lovász Local Lemma and Acyclic Edge Coloring
On the Algorithmic Lovász Local Lemma and Acyclic Edge Coloring Ioannis Giotis,2, Lefteris Kirousis,2, Kostas I. Psaromiligkos, and Dimitrios M. Thilikos,2,3 Department of Mathematics, National & Kapodistrian
More informationTopic: Balanced Cut, Sparsest Cut, and Metric Embeddings Date: 3/21/2007
CS880: Approximations Algorithms Scribe: Tom Watson Lecturer: Shuchi Chawla Topic: Balanced Cut, Sparsest Cut, and Metric Embeddings Date: 3/21/2007 In the last lecture, we described an O(log k log D)-approximation
More informationFIRST ORDER SENTENCES ON G(n, p), ZERO-ONE LAWS, ALMOST SURE AND COMPLETE THEORIES ON SPARSE RANDOM GRAPHS
FIRST ORDER SENTENCES ON G(n, p), ZERO-ONE LAWS, ALMOST SURE AND COMPLETE THEORIES ON SPARSE RANDOM GRAPHS MOUMANTI PODDER 1. First order theory on G(n, p) We start with a very simple property of G(n,
More informationExact and Approximate Equilibria for Optimal Group Network Formation
Exact and Approximate Equilibria for Optimal Group Network Formation Elliot Anshelevich and Bugra Caskurlu Computer Science Department, RPI, 110 8th Street, Troy, NY 12180 {eanshel,caskub}@cs.rpi.edu Abstract.
More informationOptimization of Submodular Functions Tutorial - lecture I
Optimization of Submodular Functions Tutorial - lecture I Jan Vondrák 1 1 IBM Almaden Research Center San Jose, CA Jan Vondrák (IBM Almaden) Submodular Optimization Tutorial 1 / 1 Lecture I: outline 1
More informationMore on NP and Reductions
Indian Institute of Information Technology Design and Manufacturing, Kancheepuram Chennai 600 127, India An Autonomous Institute under MHRD, Govt of India http://www.iiitdm.ac.in COM 501 Advanced Data
More informationApproximation Algorithms for Asymmetric TSP by Decomposing Directed Regular Multigraphs
Approximation Algorithms for Asymmetric TSP by Decomposing Directed Regular Multigraphs Haim Kaplan Tel-Aviv University, Israel haimk@post.tau.ac.il Nira Shafrir Tel-Aviv University, Israel shafrirn@post.tau.ac.il
More informationLocality Lower Bounds
Chapter 8 Locality Lower Bounds In Chapter 1, we looked at distributed algorithms for coloring. In particular, we saw that rings and rooted trees can be colored with 3 colors in log n + O(1) rounds. 8.1
More informationLecture 6: September 22
CS294 Markov Chain Monte Carlo: Foundations & Applications Fall 2009 Lecture 6: September 22 Lecturer: Prof. Alistair Sinclair Scribes: Alistair Sinclair Disclaimer: These notes have not been subjected
More informationMonotone Submodular Maximization over a Matroid
Monotone Submodular Maximization over a Matroid Yuval Filmus January 31, 2013 Abstract In this talk, we survey some recent results on monotone submodular maximization over a matroid. The survey does not
More informationThe Turán number of sparse spanning graphs
The Turán number of sparse spanning graphs Noga Alon Raphael Yuster Abstract For a graph H, the extremal number ex(n, H) is the maximum number of edges in a graph of order n not containing a subgraph isomorphic
More informationFirst order logic on Galton-Watson trees
First order logic on Galton-Watson trees Moumanti Podder Georgia Institute of Technology Joint work with Joel Spencer January 9, 2018 Mathematics Seminar, Indian Institute of Science, Bangalore 1 / 20
More informationLecture 2: January 18
CS271 Randomness & Computation Spring 2018 Instructor: Alistair Sinclair Lecture 2: January 18 Disclaimer: These notes have not been subjected to the usual scrutiny accorded to formal publications. They
More informationOn the Complexity of the Minimum Independent Set Partition Problem
On the Complexity of the Minimum Independent Set Partition Problem T-H. Hubert Chan 1, Charalampos Papamanthou 2, and Zhichao Zhao 1 1 Department of Computer Science the University of Hong Kong {hubert,zczhao}@cs.hku.hk
More informationTopics in Theoretical Computer Science April 08, Lecture 8
Topics in Theoretical Computer Science April 08, 204 Lecture 8 Lecturer: Ola Svensson Scribes: David Leydier and Samuel Grütter Introduction In this lecture we will introduce Linear Programming. It was
More informationAn Improved Approximation Algorithm for Requirement Cut
An Improved Approximation Algorithm for Requirement Cut Anupam Gupta Viswanath Nagarajan R. Ravi Abstract This note presents improved approximation guarantees for the requirement cut problem: given an
More informationIntroduction to Algorithms / Algorithms I Lecturer: Michael Dinitz Topic: Intro to Learning Theory Date: 12/8/16
600.463 Introduction to Algorithms / Algorithms I Lecturer: Michael Dinitz Topic: Intro to Learning Theory Date: 12/8/16 25.1 Introduction Today we re going to talk about machine learning, but from an
More informationGraph coloring, perfect graphs
Lecture 5 (05.04.2013) Graph coloring, perfect graphs Scribe: Tomasz Kociumaka Lecturer: Marcin Pilipczuk 1 Introduction to graph coloring Definition 1. Let G be a simple undirected graph and k a positive
More informationLecture 11 October 7, 2013
CS 4: Advanced Algorithms Fall 03 Prof. Jelani Nelson Lecture October 7, 03 Scribe: David Ding Overview In the last lecture we talked about set cover: Sets S,..., S m {,..., n}. S has cost c S. Goal: Cover
More information11.1 Set Cover ILP formulation of set cover Deterministic rounding
CS787: Advanced Algorithms Lecture 11: Randomized Rounding, Concentration Bounds In this lecture we will see some more examples of approximation algorithms based on LP relaxations. This time we will use
More informationLecture 20: LP Relaxation and Approximation Algorithms. 1 Introduction. 2 Vertex Cover problem. CSCI-B609: A Theorist s Toolkit, Fall 2016 Nov 8
CSCI-B609: A Theorist s Toolkit, Fall 2016 Nov 8 Lecture 20: LP Relaxation and Approximation Algorithms Lecturer: Yuan Zhou Scribe: Syed Mahbub Hafiz 1 Introduction When variables of constraints of an
More informationTHE METHOD OF CONDITIONAL PROBABILITIES: DERANDOMIZING THE PROBABILISTIC METHOD
THE METHOD OF CONDITIONAL PROBABILITIES: DERANDOMIZING THE PROBABILISTIC METHOD JAMES ZHOU Abstract. We describe the probabilistic method as a nonconstructive way of proving the existence of combinatorial
More information